Single-Cell Chromatin Modification Profiling Reveals Increased Epigenetic Variations with Aging

Abstract

Post-translational modifications of histone proteins and exchanges of histone variants of chromatin are central to the regulation of nearly all DNA-templated biological processes. However, the degree and variability of chromatin modifications in specific human immune cells remain largely unknown. Here, we employ a highly multiplexed mass cytometry analysis to profile the global levels of a broad array of chromatin modifications in primary human immune cells at the single-cell level. Our data reveal markedly different cell-type- and hematopoietic-lineage-specific chromatin modification patterns. Differential analysis between younger and older adults shows that aging is associated with increased heterogeneity between individuals and elevated cell-to-cell variability in chromatin modifications. Analysis of a twin cohort unveils heritability of chromatin modifications and demonstrates that aging-related chromatin alterations are predominantly driven by non-heritable influences. Together, we present a powerful platform for chromatin and immunology research. Our discoveries highlight the profound impacts of aging on chromatin modifications.

Trial registration: ClinicalTrials.gov NCT01987349 NCT03022396 NCT03022422.

Keywords: Epigenetics; aging; cell identity; chromatin modifications; heritability; histones; immune system; mass cytometry; transcriptional noise; twins.

Figure 1

Lineage-Specific Chromatin Modification Profiles in Human Immune Cells. (A) Overview of EpiTOF platform. (B) Experimental design. Two EpiTOF panels measuring 9 classes of chromatin marks are employed to analyze PBMCs derived from a cohort containing 12 CMV-seronegative healthy subjects (bio rep 1) and an independent cohort with identical demographics (bio rep 2). (C) Cell type-specific chromatin modification profiles in major immune cell subsets. EpiTOF analysis of subjects from bio rep 1. Heatmap representation of the normalized chromatin mark levels of the indicated 40 chromatin marks (x-axis) in 11 major immune cell subsets (y-axis). The normalized mark levels are centered around the mean of total PBMCs. Minimum and maximum values are shown. Color, mean across 12 subjects; dendrograms, unsupervised clustering at both axes; diameter of circle, subject-to-subject variability measured by Inverse Simpson’s Diversity Index. (D) Validation of EpiTOF data by western blotting on sorted immune cells. Western blot analysis of the whole-cell extract from sorted immune cells using the indicated antibodies. (E) Higher expression of lysine methyltransferases EZH1 and EZH2 in common lymphoid progenitors (CLP) than in common myeloid progenitors (CMP). Gene expression analysis of EZH1 and EZH2 in CLP (blue) and CMP (red) using publicly available RNA-seq dataset GSE74246. y-axis, expression levels of the indicated genes. (F) Differential enrichment of H3K27me3 and H3K4me3 are associated with gene expression reprogramming during lymphoid and myeloid lineage commitment. Top: Venn diagrams show the numbers of overlapped genes between genes enriched with H3K27me3 (left light blue circle) or H3K4me3 (right light blue circle) in T cells over monocytes and differentially expressed genes in CLP over CMP (left pink circle, lower expression in CLP over CMP; right pink circle, higher expression in CLP over CMP). Numbers of shared genes and p values of the overlaps are shown. Bottom: gene ontology enrichment analysis using the overlapped genes from the two Venn diagrams. p values for enrichment with the indicated gene ontology terms and examples of shared genes in each Venn diagram are shown.

Figure 2

Single-Cell Chromatin Modification Profile Predicts Immune Cell Identity and Uncovers Relationships between Chromatin Marks. (A) Segregation of immune cell subsets based on chromatin modification profiles. Left, PCA of single-cell EpiTOF data described in Figure 1C (bio rep 1). Each dot represents a single cell, and each principal component depicts variations of 20 chromatin marks measured using each EpiTOF panel. Analyses using single-cell data collected by EpiTOF panel 1 (top) and 2 (bottom) are shown. Cells are color-coded by immune cell types. Right, Euclidean distances of chromatin modification profiles between the indicated immune cell subtypes. (B and C) Patterns of chromatin modifications in single-cells predict immune cell identity. ROC curves discriminate the indicated immune cell lineages from the remaining PBMC subsets using a regularized logistic regression model. The AUCs measuring the sensitivity and specificity of the separation for the indicated cell types are shown (B). The model from (B) (training dataset, bio rep 1) is applied to the validation dataset (bio rep 2) to test prediction power. ROC curves and AUCs from the validation dataset are shown (C).

Figure 3

Heterogeneity of Chromatin Modification Profiles in Lymphocytes Originated from Diverse Functional Subsets. (A) T cell-specific chromatin modification profiles. EpiTOF analysis on the same cohort as in Figure 1C (bio rep 1) focusing on T cell subsets. Heatmap representation of the normalized chromatin mark levels as in Figure 1C for the indicated 40 chromatin marks (x-axis) in 11 T cell subsets (y-axis). The normalized mark levels are centered around the mean of total CD3⁺ T cells. Minimum and maximum values of normalized mark levels are shown. The mean of each chromatin mark and T cell subset pair across 12 subjects is used for plotting. Dendrograms, unsupervised clustering; diameter of circle, subject-to-subject variability measured by Inverse Simpson’s Diversity Index. (B) Memory T cells are characterized by unique patterns of chromatin marks. Heatmap representation of the effect sizes of the levels of 40 chromatin marks in the indicated memory T cells over the naïve subsets in both biological replicates. Dendrogram at y-axis, unsupervised clustering. (C) Differential chromatin marks in regulatory T cells. Heatmap representation of the effect sizes of the levels of 40 chromatin marks in Tregs over total CD4⁺ T cells in the indicated biological replicates. (D and E) Differential chromatin marks separate CD56^bright NK cells from the CD56^dim subset. Scatter plot of single-cell EpiTOF data from NK cells (bio rep 1). Each dot represent a single NK cell plotted based on CD56 (y-axis) and CD16 (x-axis) levels. Color, PC1 computed from the 20 chromatin marks measured by EpiTOF panel 1. Density plot of the two populations segregated by MixTool (green, CD56^bright; red, CD56^dim) using CD56 level is shown (D). Heatmap representation of the effect sizes computed for the levels of the indicated chromatin marks in CD56^bright over CD56^dim subsets. Data from both biological replicates are shown (E).

Figure 4

Increased Variations in Chromatin Modification Profiles with Age. (A) Data integration for aging-related analyses. Two biological replicates are merged using empirical Bayes framework to correct for batch effect. Data collected from 9 major immune cell types in the initial EpiTOF analysis (Figure 1C and S1A; exclude total CD4⁺ and CD8⁺ T cells) are integrated with data from 11 T cell subsets (Figure 3A and S3A) to create a dataset of 40 chromatin marks in 20 immune cell subsets (800 data points) for each of the 24 subjects. (B) Increased subject-to-subject variability with aging. PCA of EpiTOF data from 24 healthy subjects described in (A) (salmon, >65 years; cyan, <25 years; circle, female; square, male). Each dot represents a single subject situated in the PCA plot where each principal component summarizes the variance of 800 data points. The percentage of variance each principal component explains is shown. (C) Aging is associated with alterations of a broad range of chromatin marks. Heatmap representation of the effect sizes of the levels of the indicated chromatin mark (x-axis) and immune cell type (y-axis) pairs in older over younger subjects. Dendrogram, unsupervised clustering.

Figure 5

Cell-to-Cell Variability of Chromatin Modification Profiles Increases with Age. (A) Aging is associated with elevated single-cell variations in chromatin modifications. T-test comparison of the CVs of 800 data points between subjects from the two age groups. x-axis, p-values; directionality, CV higher in subjects <25 years (left) or higher in subject >65 years (right); color, FDR (dark blue, <5%; orange, 5–10%; light blue, >10%). Test results are ranked by FDRs from top to bottom. (B) H3K27me2, H3K27me3 and H2AK119ub show the greatest increase in cell-to-cell variability with age. The numbers of immune cell types (y-axis) in which the indicated chromatin marks (x-axis) show statistically significant (FDR<5%) increase in CVs in older subjects. Red, chromatin marks generated by PRC1 or PRC2; gray, chromatin marks that are not catalyzed by PRCs. (C) H3K27me3-marked genes are associated with increased cell-to-cell transcriptional variability with age. Changes in transcriptional variability (ΔCV) between naïve CD4⁺ T cells purified from old and young M. musculus domesticus of the indicated gene subsets are computed. Transcription profiling data and lists of genes occupied by H3K27me3 or H3K4me3 are derived from publicly available scRNA-seq (Martinez-Jimenez et al., 2017) and ChIP-seq (Wei et al., 2009) datasets, respectively. Transcriptome-wide increased transcriptional noise with age is shown. p-values, Wilcoxon’s paired rank sum test.

Figure 6

Non-Heritable Influences Explain Most Variations in Chromatin Modification Profiles. (A) Overview of heritability analyses. 9 pairs of monozygotic (MZ) twins and 10 pairs of dizygotic (DZ) twins are subject to EpiTOF analysis utilizing four EpiTOF panels covering major immune cells and T cell subsets. (B) Variance in chromatin modification profiles is largely driven by unique environmental factors. Heatmap representation of the proportions of variance explained by additive genetics (left), common environment (middle) or unique environment (right) for the indicated chromatin mark and immune cell subset pairs. Chromatin marks are ranked from top to bottom based on the average influences from additive genetics. Immune cells are ranked from left to right by the averages of additive genetics influences across all chromatin marks. The average influences of each component on all 800 data points are shown. (C and D) Aging is associated with divergent chromatin modification profiles between twins. PCA of younger (cyan) and older (salmon) twin subjects. Each dot represents a single twin subject and the twins are connected. Principal component, variance of 800 data points. The percentage of variance explained by each principal component is shown (C). Euclidean distances of 800 chromatin mark and cell type pairs are computed for each pair of twins (left) or randomly selected genetically unrelated subjects (right) from separate age groups (cyan, younger subjects; salmon, older subjects). p values for the statistical significance of increased Euclidean distance in older pairs are shown (D). (E) Concordance of chromatin modification profiles in younger MZ twins diminishes with age. Euclidean distances calculated from the 800 data points between MZ (left), DZ (middle) twins and randomly paired genetically unrelated individuals (right). Top, younger subjects; bottom, older subjects. p values for the statistical significance of distinct Euclidean distance are shown. (F) Non-heritable influences drive the increased variability in chromatin modifications with age. Spearman’s rank correlations of 40 chromatin marks across 20 immune cells types are computed for younger and older MZ twins. Each dot represents a chromatin mark. x-axis, correlation between older twin pairs; y-axis, correlation between young twin pairs. Dashed line, equal correlation in younger and older MZ twins. Arrows indicate the directions of higher concordance in younger pairs (upper left) or in older pairs (lower right).